Closed neubig closed 5 years ago
Yes! The issue is the original ELAN file that I have been working with. You should be getting a warning that it has unknown spec -- hence it cannot be properly written out after the additional tier has been added.
If you do check the corpus, the annotation is there -- it's the writing it out with pympi that fails.
You had some other ELAN files to play with, if I recall correctly?
Cool! It kind-of works with the new file.
I'm not sure why VAD needs "en" as the input tier, as specified. Shouldn't it be just the wav file?
Also, the "VAD" output is a bit weird, looking like the following screenshot. Does this look the same as what you got or is it different in my environment than yours?
It does not need 'en' as an input tier, it is indeed only using the wav file. I made the necessary changes so that the --input_tiers flag is optional, see the updated README.
I also updated the VAD model, and changed the VAD annotations into SpanTextAnnotations, marking the returned active regions. However, my ELAN cannot open the resulting output .eaf file, can you see if you can open it? It seems to be ok when I read it through a text editor :/
My version (5.5) opened it just fine actually:
Maybe we can close the issue?
In the most recent branch (specifically,
add-setup
here https://github.com/neulab/cmulab/pull/9) I tried to run the elan client, but it doesn't seem to write anything to the output directory:Is this the same behavior that you get @antonisa ?